Deciphering chemotaxis pathways using cross species comparisons
- Rebecca Hamer†1, 2,
- Pao-Yang Chen†4,
- Judith P Armitage2, 3,
- Gesine Reinert1, 2 and
- Charlotte M Deane1, 2Email author
© Hamer et al; licensee BioMed Central Ltd. 2010
Received: 10 July 2009
Accepted: 11 January 2010
Published: 11 January 2010
Chemotaxis is the process by which motile bacteria sense their chemical environment and move towards more favourable conditions. Escherichia coli utilises a single sensory pathway, but little is known about signalling pathways in species with more complex systems.
To investigate whether chemotaxis pathways in other bacteria follow the E. coli paradigm, we analysed 206 species encoding at least 1 homologue of each of the 5 core chemotaxis proteins (CheA, CheB, CheR, CheW and CheY). 61 species encode more than one of all of these 5 proteins, suggesting they have multiple chemotaxis pathways. Operon information is not available for most bacteria, so we developed a novel statistical approach to cluster che genes into putative operons. Using operon-based models, we reconstructed putative chemotaxis pathways for all 206 species. We show that cheA-cheW and cheR-cheB have strong preferences to occur in the same operon as two-gene blocks, which may reflect a functional requirement for co-transcription. However, other che genes, most notably cheY, are more dispersed on the genome. Comparison of our operons with shuffled equivalents demonstrates that specific patterns of genomic location may be a determining factor for the observed in vivo chemotaxis pathways.
We then examined the chemotaxis pathways of Rhodobacter sphaeroides. Here, the PpfA protein is known to be critical for correct partitioning of proteins in the cytoplasmically-localised pathway. We found ppfA in che operons of many species, suggesting that partitioning of cytoplasmic Che protein clusters is common. We also examined the apparently non-typical chemotaxis components, CheA3, CheA4 and CheY6. We found that though variants of CheA proteins are rare, the CheY6 variant may be a common type of CheY, with a significantly disordered C-terminal region which may be functionally significant.
We find that many bacterial species potentially have multiple chemotaxis pathways, with grouping of che genes into operons likely to be a major factor in keeping signalling pathways distinct. Gene order is highly conserved with cheA-cheW and cheR-cheB blocks, perhaps reflecting functional linkage. CheY behaves differently to other Che proteins, both in its genomic location and its putative protein interactions, which should be considered when modelling chemotaxis pathways.
Chemotaxis is the process by which motile bacteria move towards more favourable conditions by sensing their chemical environment. It is of significant medical interest, as many pathogenic bacteria depend on chemotaxis and motility to invade their hosts. For example, Helicobacter pylori, which colonizes the mucus lining of the stomach, has a chemotactic response to gastric mucin . Biofilm development depends on chemotaxis, and their formation in the lungs of cystic fibrosis patients and on medical implants can have serious consequences [2, 3]. Chemotaxis is also essential for symbiotic associations of bacteria, for example the colonization of wheat roots by the nitrogen fixing bacterium Azospirillum brasilense . In addition, chemotaxis is the canonical system used to study signalling pathways in systems biology, due to its relative simplicity for modelling purposes, and the ease with which it can be studied experimentally. A detailed, quantitative understanding of chemotaxis pathways would pave the way for the study of other, more complex signalling systems.
E. coli chemotaxis is a relatively simple biological system which is conserved across many bacterial and archaeal species, and has inspired the development of numerous mathematical models of chemotaxis [6, 7]. However, some bacteria have multiple homologues of the Che proteins which probably form more than one chemotaxis pathway. Rhodobacter sphaeroides is a well studied example of a species with multiple chemotaxis pathways. It has several homologues of E. coli CheA, B, R, W and Y proteins but none of CheZ. There are three operons encoding the majority of the chemotaxis genes (CheOp1, CheOp2 and CheOp3), as well as other unlinked loci encoding chemoreceptors and CheY homologues . CheOp2 and CheOp3 are essential for chemotaxis in the laboratory, while the physiological role of CheOp1 has not yet been established [9–13]. Of the 13 chemoreceptor homologues, 4 lack transmembrane regions and are referred to as transducer-like proteins (Tlps). These are cytoplasmic and sense the metabolic state of the cell rather than the exterior environment. Studies have shown the presence of two discrete protein clusters in the cell, with the proteins encoded by CheOp2 localising with the MCPs at the cell pole, and proteins encoded by CheOp3 localising with cytoplasmic Tlps . Chemotaxis requires signals from both these clusters to be integrated to produce a response at the single flagellar motor. Exactly how these clusters are formed and targeted to the correct position in the cell is still the subject of active research , but it is known that the PpfA protein, which is homologous to bacterial type I DNA partitioning factors and is encoded in CheOp3, is critical for correct partitioning of the cytoplasmic protein clusters on cell division .
CheA has 5 domains designated P1-P5 [17, 18]. P1 is the histidine-containing phosphotransfer (HPt) domain, P2 contains the binding site for CheY and CheB, P3 is the dimerization domain, P4 is the kinase domain which phosphorylates a conserved histidine in P1, and P5 binds to CheW and the receptors. R. sphaeroides contains two classical CheA homologues, and two with missing domains. CheA3 has only P1 and P5 domains, separated by a 794 residue long linker region which includes a CheY6-specific phosphatase domain , and CheA4 has only domains P3, P4 and P5 . The reason for the presence of such unusual homologues and their prevalence among other bacterial species is currently unknown.
Activated CheA transfers the phosphoryl group from its P1 domain to the response regulator, CheY. CheYs belong to a group of proteins termed single domain response regulators (SDRRs), but differ from classical SDRRs as they do not include an output domain. Genomes can encode a large number of SDRRs and their annotation as CheY-encoding genes usually depends on their genomic context. R. sphaeroides has at least 6 CheY homologues and a minimum of two are necessary for chemotaxis; CheY6 and either of CheY3 or CheY4. However, although all of CheY1-6 can bind to FliM, only CheY6 is capable of stopping the flagellar motor, the mechanism by which swimming direction is changed in this species . CheY6 is also unusual in that it autodephosphorylates ten times faster than CheY1-5 and E. coli CheY . It is also possible that CheY1-5 act as phosphate sinks to aid signal termination [22, 23], and/or they may compete with CheY6 for binding to FliM . Although CheY6 is the major motor-binding CheY, and is annotated as CheY because of its genome context, it is in fact more similar to the REC domain of E. coli CheB than to E. coli CheY.
While chemotaxis requires Che proteins, not all Che protein homologues are specialised for chemotaxis. Some species, such as Pseudomonas aeruginosa and Myxococcus Xanthus, contain operons encoding Che protein homologues which have become specialised for an alternative function such as twitching motility or controlling gene expression during development [24–26].
This work uses operon information to predict chemotaxis pathways from the genetic sequence of a species. Unfortunately, experimentally-based information about operons is not readily available for a large number of bacterial genomes, so here we use a novel statistical procedure based on the proximity of genes on the genome to predict clusters of che genes. These clusters are referred to from here on as operons, but it should be recognised that these are computational predictions. Previous approaches use cut-offs on operon lengths , or the distance between two adjacent genes [28, 29]e.g. the Gene Gap Method . However, methods based on operon lengths could potentially result in the distance of adjacent genes in an operon being larger than the distance between two operons. On the other hand, methods based on the distance between two adjacent genes may result in large operons in species with short genomes. In both approaches, the selection of cut-offs is essential, as one fixed cut-off may not apply to all genomes, or even to all operons, due to the large variety of genome sizes across species. Our method overcomes these problems using a multivariate normal clustering method based on the Akaike Information Criterion (AIC) .
In building our models of chemotaxis pathways, we assume that CheB, CheR, CheW and CheY are all functionally linked to CheA. We refer to all these functional linkages as 'Putative Interactions' (PIs) for simplicity. Physiologically, CheA directly interacts with CheW, CheB and CheY and has an indirect interaction with CheR, via the chemotaxis receptors. Two putatively interacting Che proteins may be encoded by genes either in the same operon, which we call a within-operon PI, or by genes in two different operons, which we call an across-operon PI. Note that the term 'interaction' does not imply a gene-gene interaction, but rather is shorthand for the fact that the encoded proteins are functionally linked and putatively interact. To date, there is still very little information about cross-talk between chemotaxis proteins encoded by different operons, so we consider five assumptions to construct parsimonious models for chemotaxis pathways from our predicted operons: (i) All Che proteins are part of a complete set required to operate chemotaxis i.e. all Che proteins are part of a functioning chemotaxis pathway. This is an idealised assumption, as some Che proteins may serve alternative functions. (ii) The proteins belonging to one pathway 'attempt' to be distinct from those belonging to other pathways. This is supported by the known pathways observed in species such as P. aeruginosa  and also allows simpler controls within a pathway. (iii) PIs tend to happen between proteins encoded within the same operon in preference to between proteins encoded in different operons, as genes within the same operon are in close proximity and are co-transcribed from the same promoter. (iv) The ranking of probabilities of within-operon PIs and of across-operon PIs is identical in every species, since Che proteins maintain the same functions across species. This may reflect horizontal gene transfer of whole pathways between species. Note that the probabilities themselves do not have to be identical, only the ranking. (v) A chemotaxis pathway tends to minimise the number of operons its proteins are encoded in. This is based on the biological conjecture that the fewer operons used, the simpler the control mechanism. Using our models, chemotaxis pathways are predicted for each of the 206 species.
We demonstrate that the organisation of genes into operons is not arbitrary, and observe that cheA and cheW are frequently adjacent within operons, as are cheB and cheR. This may reflect the close relationship of the encoded proteins, for example in forming a protein complex or in the adaption mechanism, respectively, and possibly the need for a strict stoichiometric relationship . We find that the distribution of cheY in operons is different from the other che genes and that CheY PI behaviour is different to that of other Che proteins. This finding is in line with those of Wuichet et al. , where the problem of distinguishing CheY proteins in a set of stand-alone REC domains is discussed. As CheY proteins have been shown to regulate both flagellar and pili-based motility, Wuichet et al. also argue for a special treatment of CheY. With the caveat that identifying a CheY homologue as being involved in a chemotaxis pathway is problematic, models of chemotaxis which treat CheY differently to the other Che proteins may therefore be most appropriate. We also observe that the presence of ppfA homologues within che operons is widespread, suggesting that cytoplasmic clusters of Che proteins may be common to many bacterial species. In species with multiple Che homologues, grouping of che genes into operons and localization of Che proteins into clusters in the cell are likely to be the major factors determining separation of chemotaxis pathways. Finally, our predictions of chemotaxis pathways not only closely match the available interactions reported in the literature [5, 8, 32, 35] but they also suggest pathways for hitherto less studied species.
Protein sequences in Fasta format from all 833 complete bacterial and archaeal genomes available at the NCBI  in February 2009 were downloaded and a non-redundant set of 523 genomes was created by removing multiple strains of the same species.
BLAST filtering criteria
Query sequence length
Max hit length
Domains in E. coli from CDART
BLAST 'query from' criteria
BLAST 'query to' criteria
>340 & <400
>475 & <535
>480 & <540
>90 & <160
>130 & <190
>50 & <110
>60 & <210
>120 & <200
>140 & <210
Distribution of che genes across all species studied
Number of che genes
Number of species
Isolated che genes*
Species with isolated che genes
Genomic locations of Che protein homologues in each species were found by searching the bacterial ptt files  for corresponding gi accession numbers. Of the 485 species which had a BLAST hit to at least one of CheA, B, R, W, Y, V or Z, che genes in 474 species could be located in ptt files (Additional file 1).
206 of these 474 genomes were found to contain at least one homologue of each of the core E. coli chemotaxis genes (encoding CheA, B, R, W and Y). The che genes from these genomes were grouped into operons using the statistical approach described next. All statistics are based on this data set.
Assigning genes to operons
For most organisms, operon information is not available. Hence a standard statistical clustering approach for assigning genes to operons is employed, see . Here, there is biological information to be taken into account. The candidate operons are generated from genes which are on the same DNA strand with no intervening che genes transcribed in the opposite direction . Intuitively, the maximal distance between two adjacent genes in the same operon should be smaller than the minimal distance between any two adjacent operons; this is made a requirement for the clustering algorithm. Noting that the largest gap between che genes in the E. coli chemotaxis operon is ~3,500 base pairs, the clustering algorithm assumes that when two che genes are within 3,500 base pairs, then these two genes are in the same operon. As the shortest distance between two genes from different che operons in R. sphaeroides is ~100,000 base pairs, adjacent genes are separated into two operons if their distance is greater than 100,000 base pairs. The resulting assignment of genes to operons is robust with respect to changes of this parameter (100,000 base pairs) within a reasonable range (data not shown).
The coarse assumptions are that genes within an operon have a normally distributed location across the genome, with operon-specific means. The estimation of the number of operons and their gene contents then follows analogously from estimating the number of components in a mixture of normal variables, and assigning the observations to different components of this mixture.
A statistical heuristic assumes that the position of genes, given by their centre points, are approximately independent and normally distributed with means depending on the (unknown) operon they belong to, but with the same, unknown, variance . This is an approximation based on the fact that genes are short in comparison with the length of the genome.
For fixed k, an assignment of genes into operons is chosen which minimises W k . The larger a k is chosen, the smaller W k will tend to be, but also the explanatory power of the assignment will be reduced. The optimal number of operons is deemed to be the number k which minimises the Corrected Akaike Information Criterion (AICC) , given by 2(logWk) + 2nk/(n-k+1). The additive factor 2nk/(n-k+1) penalises for choosing a large k. The thus predicted operon assignments for each species are given in Additional file 2.
In order to test whether the operon content is informative, we generated shuffled operons. Here the number and size of operons and the number of che genes are kept constant, but the che genes are shuffled, so that the allocation of che genes to operons is randomised.
Calculating gene functional order from operons
Models of chemotaxis pathways
Parsimonious models for reconstructing chemotaxis pathways are now built by predicting PIs among the multiple Che proteins in a species. In order to build these models, the diversity of patterns of che genes in operons is reduced by recording only the types of che genes; multiple occurrences of identical che genes in the same operon are counted only once.
Among the 1419 operons in the study, 85.7% do not contain multiple occurrences of che genes. By far the most frequent multiply occurring gene among the 203 operons with multiple occurrences is cheY, with 147 double occurrences, 14 triple occurrences, and 2 quadruple occurrences. Next is cheW with 42 double occurrences, then cheA and cheR both with 3 double occurrences. cheB has 1 double occurrence. See Additional file 2 for details.
To model chemotaxis pathways we look at the organisation of che genes into operons. PIs between the CheA, B, R, W and Y proteins are predicted using the operon location of the respective genes. In the base model, Model ABRWY, we simultaneously considers 4 types of protein PIs; A~B, A~R, A~W and A~Y. If there is only one CheA homologue in a species, the model assigns PIs between this CheA and all other Che proteins in that species. When there are multiple CheA homologues in a species, we assume that PIs are more likely to be between proteins encoded by genes in the same operon (within-operon) than between proteins encoded by genes from different operons (across operon). We first assign within-operon PIs between CheA and all other Che proteins encoded in the same operon (pseudo-algorithm Step 1). If the resulting pathway is not complete, i.e. if it lacks sufficient Che proteins to form a complete set (ABRWY), then cross-talk with proteins encoded in other operons is required (pseudo-algorithm Step 2). In this case incomplete operons which do not yet have edges to other operons are preferred, so that each encoded Che pathway is as distinct as possible. During the reconstruction of pathways, multiple configurations are possible. For example, an incomplete operon may connect to one of several operons having the complementary che genes (complementary operons). The preferred configuration is selected based on the operons predicted in all 206 species (Additional file 2), assuming that cross-talk between two operons is more likely if these operons are frequently observed together in many species. Finally, any non CheA-encoding operons are assigned PIs to CheA-encoding operons according to the same frequency principle (pseudo-algorithm Step 3).
Step 1: Assigning within-operon PIs
For each CheA-encoding operon
Assign within-operon PIs: CheA interacts with all other Che proteins encoded in the operon
Step 2: Completing pathways based on CheA-encoding operons
For each CheA-encoding operon
IF complete pathway THEN proceed to next operon
ELSE assign across-operon PIs to an operon which is exactly complementary
IF complete pathway THEN proceed to next operon
ELSE assign cross-talks to multiple complementary operons
IF complete pathway THEN proceed to next operon
ELSE assign cross-talks to partner operons which are not CheA-encoding
IF complete pathway THEN proceed to next operon
ELSE assign across-operon PIs to another CheA-encoding operon
Step 3: Connecting non CheA-encoding operons
For each incomplete operon which does not yet have edges to other operons
Assign PIs to CheA-encoding operon based on the operon co-occurrence frequency across all species
As the distribution of cheY in predicted operons is different from the other che genes, variants of the above model are devised to explore the possible different PI behaviours of Y. These variant models are abbreviated ABRW+Y, ABRWY+Y, and ABRWY+Y'. Model ABRW+Y considers only three types of PIs, A~B, A~R and A~W in the first step. It follows the same steps to assign protein PIs as the base model, except that Y is excluded; CheY is included in operon prediction, but is ignored when step 3 of the pseudo-algorithm is reached and are connecting operons based on the frequency principle. After the PIs A~B, A~R and A~W are assigned, every Y is connected to every A. The third model, ABRWY+Y, assigns PIs in the same way as the base model, that is by simultaneously considering the four types of PIs A~B, A~R, A~W and A~Y. In addition, the model assigns PIs between every Y and every A in the last step. The two models ABRW+Y and ABRWY+Y therefore propose that a Y can interact with all As. The fourth model, ABRWY+Y', also assigns PIs in the same way as the base model ABRWY in the first step. Then only isolated Ys, i.e. Ys not previously connected to any A, are assigned PIs with every A. It is to be emphasised that these models are based on a parsimonious approach, trying to find a simple model which explains a good amount of the observations. There are a number of exceptions from these simple rules.
Known interaction data for E. coli CheY was obtained from the Database of Interacting Proteins (DIP) . Protein disorder was predicted using RONN . T-Coffee  and Bl2seq  were used for sequence alignment.
Results and Discussion
206 species were found to have at least one homologue of each of the five core Che proteins (CheA, B, R, W and Y) and 61 (30%) of these have more than one of each, suggesting the existence of multiple chemotaxis pathways in many species.
These 206 species were briefly examined for evidence of flagellar gene homologues to determine whether they are likely to be motile. BLAST searches using the E. coli proteins FliC, FliD, FliF, FLiG, FliM, FliN, MotA and MotB were carried out and all hits with an e-value of 10 or less were accepted. 147 species had hits to all of these proteins. 18 species had hits to none of them, but according to HAMAP  only 4 of these have no flagella (Candidatus Methanoregula boonei, Desulfococcus oleovorans, Thermococcus onnurineus and Trichodesmium erythraeum). However, Thermococcus onnurineus is an archaea and KEGG  indicates it does have flagella-related proteins present. Trichodesmium erythraeum is a cyanobacterium and therefore might move using pili, as may Candidatus Methanoregula boonei and Desulfococcus oleovorans. We therefore conclude that the 206 species are all likely to be motile rather than using their Che proteins only for other functions. Singh et al.  also found that the majority of the species in their study which encode genes for CheA, B, R, W and Y were annotated as being motile.
Organisation of che genes in the top 10 most frequently found operons
Number of operons
Relative frequency (%)
Number of species
Relative frequency (%)
Distribution of che genes within operons and within species
Number of homologues
Number of operons
Number of homologues
Number of species
Isolated che genes, complete operons and complementary operons
The order of genes reflects their functional mechanism
Relative occurrence of ordered gene pairs
1 st gene
2 nd gene
1 st gene
Close proximity of genes may indicate that the proteins they encode are co-localised to form a complex, aided by co-transcription and co-translation. Dandekar et al. found that conservation of gene order was a 'fingerprint' of proteins which physically interact . CheA and CheW have been shown to interact in vitro and in vivo [53, 54], and seen to interact in a crystal structure (PDB identifier 2CH4). It has also been shown that CheA and CheW co-localise to the MCPs in E. coli [56, 57]. However, no interaction between CheR and CheB has been shown. While CheR and CheB co-localise to the transmembrane receptors in E. coli [5, 58], in R. sphaeroides, CheR2 and R3 are localised to the cytoplasmic and membrane bound chemosensory clusters respectively. However, CheB1 and B2 are diffuse in the cytoplasm, making the role of localisation unclear.
Comparison to known chemotaxis operons
51% of the 206 species studied and 82% of the 61 species with putative multiple pathways have at least one complete chemotaxis operon (containing cheA, B, R, W and Y). In order to ascertain if these operons may encode similar pathways to those identified in R. sphaeroides, we examined whether known non che genes are also found within the operons. The proteins encoded in R. sphaeroides CheOp3 localise in cytoplasmic clusters with Tlp receptors. PpfA, encoded in CheOp3, is known to be critical for correct partitioning of these protein clusters upon cell division . However, PpfA is a ParA homologue, and ParA-ParB pairs may be involved in DNA segregation upon cell division . We found that 77 species (37%) have a PpfA homologue encoded in a che operon without a ParB homologue being encoded in the same operon (Additional file 5). 86% of these species also have a putative cytoplasmic chemotaxis receptor homologue. Strikingly, 61% of species that potentially have multiple chemotaxis pathways have a PpfA homologue encoded in a che operon, and of these 97% also have a putative cytoplasmic chemotaxis receptor homologue. The fact that separate groups of che genes are found so often on bacterial genomes and that ppfA, a non che gene, is often present in these operons suggests that not only is the grouping of che genes a common way of separating chemotaxis pathways but that organised cytoplasmic clusters of Che proteins may be present in a significant number of species.
Prediction of chemotaxis pathways
We predict chemotaxis pathways in all 206 species using the 4 models described in the methods. When building these models, we ignore multiple occurrences of che genes in an operon, as we assume that each copy of a che gene in an operon will have the same interaction behaviour. Of the 1419 operons analysed, 203 (14%) have multiple occurrences of one or more che genes. The most frequently observed che gene to have multiple copies within a che operon is cheY, with multiple copies occurring in 164 operons (Additional file 6). Our 4 models explicitly consider different possible PI behaviours of cheY. The other four che genes, A, B, R and W, occur as multiple copies in only 3, 1, 3 and 42 operons respectively.
We calculate the occurrences of each type of PI (Additional file 7) for all 4 models. Due to the fact that multiple copies of che genes in operons are ignored, recording an occurrence is to be understood as "there is at least one PI of this type" taking place. For example, in the operon AWW we count one within-operon A~W PI. With this interpretation, our conclusions are not affected by multiple occurrences of a gene within an operon. As Che proteins may take on multiple functional roles, it is to be noted that ignoring multiplicity of copies may have resulted in neglecting some interesting phenomena. This decision was reached not only for the sake of parsimony, but also due to the lack of information about these multiple roles and their effects on chemotaxis pathways. We also do not explicitly consider the multiple domain organization of proteins in our models. In the case of the five-domain protein CheA, we put this protein in a unique position such that all other proteins interact with it directly or indirectly. As a multiple domain protein, CheA has an central role in our models (Figure 1). It has PIs to CheW, CheY, CheB and CheR, which may reflect the connectivity from its multiple domains.
The chi-square tests of homogeneity (Additional file 8) show that models ABRWY, ABRWY+Y and ABRWY+Y' have similar PI patterns for A~B, A~R and A~W, whereas these PIs in model ABRW+Y are significantly different from the other models. When the A~Y PI is included, all models are significantly different, except ABRWY and ABRWY+Y'. The PI behaviour of CheY is still not fully understood, so all three possible PI behaviours of CheY discussed here are plausible.
PI behaviour of Che proteins
The predicted PIs from each model are then compared to those derived using shuffled operons (Additional file 9 and 10). In all 4 models the predicted PIs are significantly different from shuffled PIs (all p-values < 0.0001, chi-square tests of homogeneity) suggesting again that the organisation of che genes into operons is not arbitrary.
Ranking of the relative occurrence of PIs in predicted pathways
Model ABRWY+Y' is the most parsimonious of our models satisfying our assumptions. The ranking of the relative frequencies of PIs shows that all within-operon PIs rank higher than cross-operon PIs (Table 6 and Additional file 10), unlike for models ABRW+Y and model ABRWY+Y. Figure 5 compares the relative occurrence of PIs in the predicted pathways to their relative occurrence in randomly shuffled pathways, where the che genes are assigned to operons at random. For all PIs, the within-operon relative occurrence in the predicted pathways are significantly higher than the occurrence using shuffled pathways. For A~B, A~R and A~W, the across-operon relative occurrence of PIs in the predicted pathways are significantly lower than in randomly shuffled pathways. This finding is consistent with our model assumption that within-operon PIs are used in preference to across-operon PIs. In contrast, for A~Y, whether the across-operon PIs yield a higher or lower relative occurrence compared to randomly shuffled pathways is model dependent. In model ABRWY there is no significant difference. In model ABRWY+Y', across-operon occurrence of PIs in the predicted pathways are rarer than in shuffled pathways, but in models ABRW+Y and ABRWY+Y, the predicted pathways result in more occurrence of across-operon PIs than shuffled pathways. This finding suggests that the A~Y PI behaves differently to the other PIs, and the models in which Y is treated differently to other Che proteins may be more appropriate than Model ABRWY. For all models, across-operon A~Y PIs occur with much higher frequency than A~B, A~R and A~W across-operon PIs. This supports our finding that isolated CheYs, though not different in sequence to those encoded within che operons, may differ in their expression patterns, and hence their ability to interact in vivo.
CheY may have additional functions
The CheY-like receiver domain (REC domain) is a common regulatory module in many bacterial proteins . It is frequently found in association with DNA-binding domains but is also found as a domain in other proteins, such as in CheB, and can function alone as anSDRR. Distinguishing cheY genes from those encoding non-CheY SDRRs has so far proved impossible from sequence alone, hence the high frequency of isolated 'cheY' genes may be spurious. However, it may also imply that CheY has additional functions and interacts with proteins other than those in chemotaxis pathways. Further evidence for this comes from the 50 species found which have at least 1 CheY homologue but no other Che proteins (A, B, R, W, V or Z), and from the Database of Interacting Proteins (DIP) which suggests that E. coli CheY interacts with the pyruvate dehydrogenase complex (PDHc). PDHc ultimately causes formation of acetyl-CoA, and thus possibly CheY can autoacetylate with acetyl-CoA as the acetyl donor; it is known that acetylation of CheY can activate it and can generate clockwise flagellar rotation . In addition, CheY6 from R. sphaeroides, discussed shortly, may be a paradigm for a specific type of CheY which may be capable of binding to multiple ligands.
Comparison of predicted PIs to interactions reported in the literature
We compare our predicted pathways to those interactions reported in the literature. The prediction for E. coli is straightforward as the pathway only involves one operon, and our predictions, assigning all within-operon PIs, are verified  (Additional file 11). Similarly, the predictions are verified for other organisms with only one complete operon, such as Salmonella enterica serovar Typhimurium and Sinorhizobium meliloti .
P. aeruginosa has 5 operons containing homologues of E. coli chemotaxis proteins, designated as Clusters I-V. Our four models correctly predict all known PIs, including the across-operon PI between A (Cluster I) and R (Cluster V) (Additional file 12). The models also predict unreported PIs in Cluster II. The predicted pathways coincide with the previous finding that Cluster II functions separately from Cluster I .
Identifying pathways in R. sphaeroides using sequence information
As a potential method to improve our models, we examined the Che homologues in R. sphaeroides to see if it was possible to identify to which pathway a Che homologue would belong based on sequence level properties. However, this proved extremely difficult. For example CheW2 and CheW3, which are present in the same pathway, are no more similar to each other than to the other CheW homologues (Additional file 13). Even residues thought to be involved in contacts to CheA are not conserved between homologues from the same gene group. Further examination using the Evolutionary Trace method , where both sequence and structure are considered, also revealed no patterns of conservation. In addition, regions of the membrane-bound and cytoplasmic chemotaxis receptors putatively used for binding to CheA/CheW  are found to be extremely similar (Additional file 14), suggesting that both these types of receptors could bind to all the CheA/CheW homologues in vitro. It is also known that CheY5, encoded by CheOp1, can restore chemotaxis in a CheY3/CheY4 deletion mutant when expressed from a plasmid (unpublished data - JPA), and CheA2 has been shown to phosphorylate all CheY homologues in R. sphaeroides in vitro . Given this evidence, we propose that localization of proteins into distinct clusters in the cell, based on their operon groupings, is likely to be the key determinant separating pathways in vivo.
The Non-classical CheA homologues of R. sphaeroides are rare
CheA in E. coli is made up of 5 domains (P1 to P5). In R. sphaeroides, CheA3 only has two of these domains, P1 and P5, connected by a long linker. This architecture was not found repeated in any of the complete genomes searched, and an online search of CDART  revealed a similar protein only in the related Roseovarius sp., Caldicellulosiruptor saccharolyticus, Desulfuromonas acetoxidans and Anaerocellum thermophillum. However, the linker region between the two domains is considerably shorter than that in CheA3 for all the but the protein in Roseovarius sp. In R. sphaeroides there is a second non-classical CheA, CheA4, which has only P3, P4 and P5 domains. Only 4 other species examined had such a CheA homologue (Caldicellulosiruptor saccharolyticus, Thermoanaerobacter tengcongensis, Salinibacter ruber and Agrobacterium vitis). These non-classical CheA homologues are therefore apparently very rare, and as such are probably not a useful paradigm for modelling, although sequencing of more bacterial genomes may reveal other such proteins in the future.
CheY6 from R. sphaeroides
The aromatic residue Y106 is known to be involved in E. coli CheY function. However, CheY6 lacks an equivalent residue (Figure 7A), yet it is the only CheY homologue in R. sphaeroides which is able to stop the flagellar motor . We also predict that CheY6 has a disordered C-terminal region not seen in the other CheY homologues (Figure 7B). The combination of these differences may help to explain why CheY6 is found to auto-dephosphorylate ten times faster than CheY1-5. The flexible, disordered region may allow CheY6 to bind to multiple ligands. In order to ascertain whether there are CheY6-like homologues in other species, the extent of disorder in CheY homologues across all species was examined. 103 species (50% of all species studied) were found to have a CheY homologue with significant disorder present at the C-terminus. Of the 207 CheY homologues with C-terminal disorder (Additional file 16), 118 (from 77 species) were found in a che operon, as is CheY6 from R. sphaeroides, and 89 (from 60 species) were isolated. 37 species have a CheY homologue which has C-terminal disorder and is also lacking the important aromatic residue. 52 of the 61 species with more than one of each Che protein (85%) have a CheY homologue with C-terminal disorder, and 19 of these species have a CheY-homologue which is also lacking the aromatic residue. This suggests that CheY6 may be a common type of CheY, but must be identified by the presence of a disordered C-terminus combined with the lack of aromatic residue, rather than by sequence searches alone.
Bacterial chemotaxis is widely used in systems biology as a paradigm for signal processing. If this system can be fully understood, it would provide a basis for understanding other, more complex signalling systems. Chemotaxis in E. coli, and a few other species such as R. sphaeroides, has been widely studied, but the extent to which chemotaxis pathways in these species are representative of bacterial chemotaxis as a whole has not yet been established. This work aims to address this by undertaking an analysis of a large, non-redundant set of complete bacterial genomes.
We show that homologues of all of the core chemotaxis proteins (CheA, B, R, W and Y) are present in many species. We developed a novel operon identifier and show that che genes tend to be grouped into putative operons, with complete operons containing all 5 che genes being found in 51% of species examined.
The existence of multiple homologues of all these proteins in 30% of the 206 species studied suggests that the presence of more than one chemotaxis pathway is relatively common, and therefore that the E. coli paradigm of chemotaxis is not appropriate for a large number of bacteria. The question then arises as to how the different chemotaxis pathways in such species are kept distinct. In R. sphaeroides, proteins involved in two different chemotaxis pathways are known to be expressed simultaneously from two separate operons, but the proteins of one pathway are localised to the cell poles and the proteins of the other to a cluster in the cytoplasm. The PpfA protein, encoded in CheOp3, is known to be critical for the correct partitioning of the cytoplasmic Che protein cluster on cell division. We found PpfA homologues within our che operons in 37% of species studied, and in 61% of species which putatively have multiple chemotaxis pathways. This suggests that cytoplasmic clusters of Che proteins may occur in many other bacteria. We propose that grouping of che genes into operons and localization of proteins into clusters in the cell are likely to be the major factors determining the separation of multiple chemotaxis pathways within a species. Chemotaxis in R. sphaeroides may therefore provide a useful model for species with multiple chemotaxis pathways. However, this species encodes some apparently non-typical chemotaxis components, CheA3, CheA4 and CheY6. We found that the variants of CheA proteins are rare in the species we examined. However, the CheY6 variant appears to be a common type of CheY, with a significantly disordered C-terminal region which may be functionally significant
The grouping of chemotaxis genes from a large number of species into putative operons allowed us to examine the general distribution of che genes in bacteria. While most che genes, particularly cheA, were usually found to be situated within che operons, the distribution of cheY is different, with isolated cheY genes being extremely common. CheY PI behaviour was also predicted to be different to that of other Che proteins, and models which take these factors into account are likely to be more realistic than those which treat all Che proteins in an identical way.
Finally, gene order in che operons was found to be important with cheA-cheW and cheR-cheB blocks observed in our data. These likely reflect functional linkage of the encoded proteins. In general, the organisation of genes into operons may provide information for the inference of gene functional order, and conserved proximity between genes may suggest that the genes are involved in similar biological mechanisms. The order of genes appears to be important at both the within-operon and between-operon levels.
The authors would like to thank Dr George Wadhams, Dr Steven Porter and Dr Mark Roberts for helpful discussion, and two anonymous referees for very helpful comments which led to an improvement of the paper.
We would like to thank the BBSRC and EPSRC for funding this work through OCISB.
- Foynes S, Dorrell N, Ward SJ, Stabler RA, McColm AA, Rycroft AN, Wren BW: Helicobacter pylori possesses two CheY response regulators and a histidine kinase sensor, CheA, which are essential for chemotaxis and colonization of the gastric mucosa. Infect Immun. 2000, 68: 2016-2023. 10.1128/IAI.68.4.2016-2023.2000PubMed CentralView ArticlePubMedGoogle Scholar
- Costerton JW, Stewart PS, Greenberg EP: Bacterial biofilms: a common cause of persistent infections. Science. 1999, 284: 1318-1322. 10.1126/science.284.5418.1318View ArticlePubMedGoogle Scholar
- Bryers JD: Medical biofilms. Biotechnol Bioeng. 2008, 100: 1-18. 10.1002/bit.21838PubMed CentralView ArticlePubMedGoogle Scholar
- Broek Vande A, Lambrecht M, Vanderleyden J: Bacterial chemotactic motility is important for the initiation of wheat root colonization by Azospirillum brasilense. Microbiology. 1998, 144 (Pt 9): 2599-2606. 10.1099/00221287-144-9-2599View ArticleGoogle Scholar
- Sourjik V: Receptor clustering and signal processing in E. coli chemotaxis. Trends Microbiol. 2004, 12: 569-576. 10.1016/j.tim.2004.10.003View ArticlePubMedGoogle Scholar
- Tindall MJ, Porter SL, Maini PK, Gaglia G, Armitage JP: Overview of mathematical approaches used to model bacterial chemotaxis I: the single cell. Bull Math Biol. 2008, 70: 1525-1569. 10.1007/s11538-008-9321-6View ArticlePubMedGoogle Scholar
- Tindall MJ, Maini PK, Porter SL, Armitage JP: Overview of mathematical approaches used to model bacterial chemotaxis II: bacterial populations. Bull Math Biol. 2008, 70: 1570-1607. 10.1007/s11538-008-9322-5View ArticlePubMedGoogle Scholar
- Porter SL, Wadhams GH, Armitage JP: Rhodobacter sphaeroides : complexity in chemotactic signalling. Trends Microbiol. 2008, 16: 251-260. 10.1016/j.tim.2008.02.006View ArticlePubMedGoogle Scholar
- Porter SL, Warren AV, Martin AC, Armitage JP: The third chemotaxis locus of Rhodobacter sphaeroides is essential for chemotaxis. Mol Microbiol. 2002, 46: 1081-1094. 10.1046/j.1365-2958.2002.03218.xView ArticlePubMedGoogle Scholar
- Hamblin PA, Maguire BA, Grishanin RN, Armitage JP: Evidence for two chemosensory pathways in Rhodobacter sphaeroides. Mol Microbiol. 1997, 26: 1083-1096. 10.1046/j.1365-2958.1997.6502022.xView ArticlePubMedGoogle Scholar
- Martin AC, Wadhams GH, Armitage JP: The roles of the multiple CheW and CheA homologues in chemotaxis and in chemoreceptor localization in Rhodobacter sphaeroides. Mol Microbiol. 2001, 40: 1261-1272. 10.1046/j.1365-2958.2001.02468.xView ArticlePubMedGoogle Scholar
- Martin AC, Wadhams GH, Shah DS, Porter SL, Mantotta JC, Craig TJ, Verdult PH, Jones H, Armitage JP: CheR- and CheB-dependent chemosensory adaptation system of Rhodobacter sphaeroides. J Bacteriol. 2001, 183: 7135-7144. 10.1128/JB.183.24.7135-7144.2001PubMed CentralView ArticlePubMedGoogle Scholar
- Ward MJ, Bell AW, Hamblin PA, Packer HL, Armitage JP: Identification of a chemotaxis operon with two cheY genes in Rhodobacter sphaeroides. Mol Microbiol. 1995, 17: 357-366. 10.1111/j.1365-2958.1995.mmi_17020357.xView ArticlePubMedGoogle Scholar
- Wadhams GH, Warren AV, Martin AC, Armitage JP: Targeting of two signal transduction pathways to different regions of the bacterial cell. Mol Microbiol. 2003, 50: 763-770. 10.1046/j.1365-2958.2003.03716.xView ArticlePubMedGoogle Scholar
- Wadhams G, Martin A, Warren A, Armitage J: Requirements for chemotaxis protein localization in Rhodobacter sphaeroides. Mol Microbiol. 2005, 58: 895-902. 10.1111/j.1365-2958.2005.04880.xView ArticlePubMedGoogle Scholar
- Thompson SR, Wadhams GH, Armitage JP: The positioning of cytoplasmic protein clusters in bacteria. Proc Natl Acad Sci USA. 2006, 103: 8209-8214. 10.1073/pnas.0600919103PubMed CentralView ArticlePubMedGoogle Scholar
- Swanson RV, Schuster SC, Simon MI: Expression of CheA fragments which define domains encoding kinase, phosphotransfer, and CheY binding activities. Biochemistry. 1993, 32: 7623-7629. 10.1021/bi00081a004View ArticlePubMedGoogle Scholar
- Bilwes AM, Alex LA, Crane BR, Simon MI: Structure of CheA, a signal-transducing histidine kinase. Cell. 1999, 96: 131-141. 10.1016/S0092-8674(00)80966-6View ArticlePubMedGoogle Scholar
- Porter SL, Roberts MA, Manning CS, Armitage JP: A bifunctional kinase-phosphatase in bacterial chemotaxis. Proc Natl Acad Sci USA. 2008, 105: 18531-18536. 10.1073/pnas.0808010105PubMed CentralView ArticlePubMedGoogle Scholar
- Ferre A, De La Mora J, Ballado T, Camarena L, Dreyfus G: Biochemical study of multiple CheY response regulators of the chemotactic pathway of Rhodobacter sphaeroides. J Bacteriol. 2004, 186: 5172-5177. 10.1128/JB.186.15.5172-5177.2004PubMed CentralView ArticlePubMedGoogle Scholar
- Porter SL, Armitage JP: Phosphotransfer in Rhodobacter sphaeroides chemotaxis. J Mol Biol. 2002, 324: 35-45. 10.1016/S0022-2836(02)01031-8View ArticlePubMedGoogle Scholar
- Shah DS, Porter SL, Harris DC, Wadhams GH, Hamblin PA, Armitage JP: Identification of a fourth cheY gene in Rhodobacter sphaeroides and interspecies interaction within the bacterial chemotaxis signal transduction pathway. Mol Microbiol. 2000, 35: 101-112. 10.1046/j.1365-2958.2000.01680.xView ArticlePubMedGoogle Scholar
- Shah DS, Porter SL, Martin AC, Hamblin PA, Armitage JP: Fine tuning bacterial chemotaxis: analysis of Rhodobacter sphaeroides behaviour under aerobic and anaerobic conditions by mutation of the major chemotaxis operons and cheY genes. Embo J. 2000, 19: 4601-4613. 10.1093/emboj/19.17.4601PubMed CentralView ArticlePubMedGoogle Scholar
- Kirby JR: Chemotaxis-like regulatory systems: unique roles in diverse bacteria. Annu Rev Microbiol. 2009, 63: 45-59. 10.1146/annurev.micro.091208.073221View ArticlePubMedGoogle Scholar
- Stover CK, Pham XQ, Erwin AL, Mizoguchi SD, Warrener P, Hickey MJ, Brinkman FS, Hufnagle WO, Kowalik DJ, Lagrou M, et al.: Complete genome sequence of Pseudomonas aeruginosa PA01, an opportunistic pathogen. Nature. 2000, 406: 959-964. 10.1038/35023079View ArticlePubMedGoogle Scholar
- Zusman DR, Scott AE, Yang Z, Kirby JR: Chemosensory pathways, motility and development in Myxococcus xanthus. Nat Rev Microbiol. 2007, 5: 862-872. 10.1038/nrmicro1770View ArticlePubMedGoogle Scholar
- Salgado H, Moreno-Hagelsieb G, Smith TF, Collado-Vides J: Operons in Escherichia coli : genomic analyses and predictions. Proc Natl Acad Sci USA. 2000, 97: 6652-6657. 10.1073/pnas.110147297PubMed CentralView ArticlePubMedGoogle Scholar
- Zaslaver A, Mayo A, Ronen M, Alon U: Optimal gene partition into operons correlates with gene functional order. Physical Biology. 2006, 3: 183-189. 10.1088/1478-3975/3/3/003View ArticlePubMedGoogle Scholar
- Overbeek R, Fonstein M, D'Souza M, Pusch GD, Maltsev N: The use of gene clusters to infer functional coupling. Proc Natl Acad Sci USA. 1999, 96: 2896-2901. 10.1073/pnas.96.6.2896PubMed CentralView ArticlePubMedGoogle Scholar
- Yan Y, Moult J: Detection of operons. Proteins. 2006, 64: 615-628. 10.1002/prot.21021View ArticlePubMedGoogle Scholar
- Akaike H: New Look at Statistical-Model Identification. IEEE Transactions on Automatic Control. 1974, Ac19: 716-723. 10.1109/TAC.1974.1100705.View ArticleGoogle Scholar
- Ferrandez A, Hawkins A, Summerfield D, Harwood C: Cluster II che Genes from Pseudomonas aeruginosa Are Required for an Optimal Chemotactic Response. J Bacteriol. 2002, 184: 4374-4383. 10.1128/JB.184.16.4374-4383.2002PubMed CentralView ArticlePubMedGoogle Scholar
- Lovdok L, Kollmann M, Sourjik V: Co-expression of signaling proteins improves robustness of the bacterial chemotaxis pathway. J Biotechnol. 2007, 129: 173-180. 10.1016/j.jbiotec.2007.01.024View ArticlePubMedGoogle Scholar
- Wuichet K, Alexander RP, Zhulin IB: Comparative genomic and protein sequence analyses of a complex system controlling bacterial chemotaxis. Methods Enzymol. 2007, 422: 1-31. 10.1016/S0076-6879(06)22001-9PubMed CentralPubMedGoogle Scholar
- Baker MD, Wolanin PM, Stock JB: Signal transduction in bacterial chemotaxis. Bioessays. 2006, 28: 9-22. 10.1002/bies.20343View ArticlePubMedGoogle Scholar
- NCBI Bacterial Genome Protein Sequences. ftp://ftp.ncbi.nlm.nih.gov/genomes/Bacteria/all.faa.tar.gz
- Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ: Basic local alignment search tool. J Mol Biol. 1990, 215: 403-410.View ArticlePubMedGoogle Scholar
- Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997, 25: 3389-3402. 10.1093/nar/25.17.3389PubMed CentralView ArticlePubMedGoogle Scholar
- Geer LY, Domrachev M, Lipman DJ, Bryant SH: CDART: protein homology by domain architecture. Genome Res. 2002, 12: 1619-1623. 10.1101/gr.278202PubMed CentralView ArticlePubMedGoogle Scholar
- Terry K, Go AC, Ottemann KM: Proteomic mapping of a suppressor of non-chemotactic cheW mutants reveals that Helicobacter pylori contains a new chemotaxis protein. Mol Microbiol. 2006, 61: 871-882. 10.1111/j.1365-2958.2006.05283.xView ArticlePubMedGoogle Scholar
- Zhulin IB, Nikolskaya AN, Galperin MY: Common extracellular sensory domains in transmembrane receptors for diverse signal transduction pathways in bacteria and archaea. J Bacteriol. 2003, 185: 285-294. 10.1128/JB.185.1.285-294.2003PubMed CentralView ArticlePubMedGoogle Scholar
- NCBI Bacterial Genome PTT files. ftp://ftp.ncbi.nlm.nih.gov/genomes/Bacteria/all.ptt.tar.gz
- Mardia KV, Kent JT, Bibby JM: Multivariate Analysis. 1979, London: Academic PressGoogle Scholar
- Hurvich C, Tsai CL: Regression and Time-Series Model Selection in Small Samples. Biometrika. 1989, 76: 297-307. 10.1093/biomet/76.2.297.View ArticleGoogle Scholar
- Dandekar T, Snel B, Huynen M, Bork P: Conservation of gene order: a fingerprint of proteins that physically interact. Trends Biochem Sci. 1998, 23: 324-328. 10.1016/S0968-0004(98)01274-2View ArticlePubMedGoogle Scholar
- Salwinski L, Miller CS, Smith AJ, Pettit FK, Bowie JU, Eisenberg D: The Database of Interacting Proteins: 2004 update. Nucleic Acids Res. 2004, 32: D449-451. 10.1093/nar/gkh086PubMed CentralView ArticlePubMedGoogle Scholar
- Yang ZR, Thomson R, McNeil P, Esnouf RM: RONN: the bio-basis function neural network technique applied to the detection of natively disordered regions in proteins. Bioinformatics. 2005, 21: 3369-3376. 10.1093/bioinformatics/bti534View ArticlePubMedGoogle Scholar
- Notredame C, Higgins DG, Heringa J: T-Coffee: A novel method for fast and accurate multiple sequence alignment. J Mol Biol. 2000, 302: 205-217. 10.1006/jmbi.2000.4042View ArticlePubMedGoogle Scholar
- Tatusova TA, Madden TL: BLAST 2 Sequences, a new tool for comparing protein and nucleotide sequences. FEMS Microbiol Lett. 1999, 174: 247-250. 10.1111/j.1574-6968.1999.tb13575.xView ArticlePubMedGoogle Scholar
- Lima T, Auchincloss AH, Coudert E, Keller G, Michoud K, Rivoire C, Bulliard V, de Castro E, Lachaize C, Baratin D, et al.: HAMAP: a database of completely sequenced microbial proteome sets and manually curated microbial protein families in UniProtKB/Swiss-Prot. Nucleic Acids Res. 2009, 37: D471-478. 10.1093/nar/gkn661PubMed CentralView ArticlePubMedGoogle Scholar
- Kanehisa M, Goto S: KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000, 28: 27-30. 10.1093/nar/28.1.27PubMed CentralView ArticlePubMedGoogle Scholar
- Singh AH, Wolf DM, Wang P, Arkin AP: Modularity of stress response evolution. Proc Natl Acad Sci USA. 2008, 105: 7500-7505. 10.1073/pnas.0709764105PubMed CentralView ArticlePubMedGoogle Scholar
- Gegner JA, Graham DR, Roth AF, Dahlquist FW: Assembly of an MCP receptor, CheW, and kinase CheA complex in the bacterial chemotaxis signal transduction pathway. Cell. 1992, 70: 975-982. 10.1016/0092-8674(92)90247-AView ArticlePubMedGoogle Scholar
- Boukhvalova MS, Dahlquist FW, Stewart RC: CheW binding interactions with CheA and Tar. Importance for chemotaxis signaling in Escherichia coli. J Biol Chem. 2002, 277: 22251-22259. 10.1074/jbc.M110908200View ArticlePubMedGoogle Scholar
- Park SY, Borbat PP, Gonzalez-Bonet G, Bhatnagar J, Pollard AM, Freed JH, Bilwes AM, Crane BR: Reconstruction of the chemotaxis receptor-kinase assembly. Nat Struct Mol Biol. 2006, 13: 400-407. 10.1038/nsmb1085View ArticlePubMedGoogle Scholar
- Maddock JR, Shapiro L: Polar location of the chemoreceptor complex in the Escherichia coli cell. Science. 1993, 259: 1717-1723. 10.1126/science.8456299View ArticlePubMedGoogle Scholar
- Sourjik V, Berg HC: Localization of components of the chemotaxis machinery of Escherichia coli using fluorescent protein fusions. Mol Microbiol. 2000, 37: 740-751. 10.1046/j.1365-2958.2000.02044.xView ArticlePubMedGoogle Scholar
- Kentner D, Sourjik V: Spatial organization of the bacterial chemotaxis system. Curr Opin Microbiol. 2006, 9: 619-624. 10.1016/j.mib.2006.10.012View ArticlePubMedGoogle Scholar
- Gerdes K, Moller-Jensen J, Bugge Jensen R: Plasmid and chromosome partitioning: surprises from phylogeny. Mol Microbiol. 2000, 37: 455-466. 10.1046/j.1365-2958.2000.01975.xView ArticlePubMedGoogle Scholar
- Galperin M, Gomelsky M: Bacterial signal transduction modules: from genomics to biology. ASM News. 2005, 71: 326-333.Google Scholar
- Barak R, Yan J, Shainskaya A, Eisenbach M: The chemotaxis response regulator CheY can catalyze its own acetylation. J Mol Biol. 2006, 359: 251-265. 10.1016/j.jmb.2006.03.033View ArticlePubMedGoogle Scholar
- Scharf B, Schmitt R: Sensory transduction to the flagellar motor of Sinorhizobium meliloti. J Mol Microbiol Biotechnol. 2002, 4: 183-186.PubMedGoogle Scholar
- Porter SL, Armitage JP: Chemotaxis in Rhodobacter sphaeroides requires an atypical histidine protein kinase. J Biol Chem. 2004, 279: 54573-54580. 10.1074/jbc.M408855200View ArticlePubMedGoogle Scholar
- Porter SL, Wadhams GH, Martin AC, Byles ED, Lancaster DE, Armitage JP: The CheYs of Rhodobacter sphaeroides. J Biol Chem. 2006, 281: 32694-32704. 10.1074/jbc.M606016200View ArticlePubMedGoogle Scholar
- Innis CA, Shi J, Blundell TL: Evolutionary trace analysis of TGF-beta and related growth factors: implications for site-directed mutagenesis. Protein Eng. 2000, 13: 839-847. 10.1093/protein/13.12.839View ArticlePubMedGoogle Scholar
- Gouet P, Courcelle E, Stuart DI, Metoz F: ESPript: analysis of multiple sequence alignments in PostScript. Bioinformatics. 1999, 15: 305-308. 10.1093/bioinformatics/15.4.305View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.